Okay, I lied. With the state of the world as it is, these are not the thoughts that keep me awake at night. Climate change, the economy, and ongoing conflicts around the world (among other things) are of more immediate concern. But if you think that AI is only a threat to creatives in the film and TV industries, I ask that you ponder these five questions:
1. If I have scripts available online (via a publisher, my own site, etc.), can they be used to train AI models? Even more concerning, have my scripts already been used in this way?
Ralph Sevush: Quite possibly. The tech companies engaged in the training of generative AI models are “scraping” the internet for every bit of digital data they can get their virtual hands on and they do not ask for permission (claiming it to be a “fair use”), nor do they identify the work they’ve taken, nor do they currently label the resulting material as being generated by AI, so it is impossible to know whether a particular writer’s work has been ingested. That’s why there are currently lawsuits challenging the right of those companies to do this without permission. In any event, writers should also be wary of publishing agreements that allow publishers to make their catalogues available to such data scrapers. In fact, even where the contract is silent on the matter, there should probably be included an affirmative prohibition against it.
2. What is stopping producers and content owners from using AI to create such things as jukebox musical librettos? They own or control the musical IP, and what they need is a script tailored around the music. Will these rights owners continue to hire playwrights to craft stories, or will they simply use AI?
Ralph Sevush: There are at least five things I can think of that are currently disincentivizing theatre producers from using AI to write theatrical work:
- Producers love theatre; that’s why they produce it. So at least some of them will find this technology to be antithetical to the humanity that theater celebrates and detrimental to the art form itself (NOTE: They might get over their reservations fairly quickly if AI-based works prove to be profitable);
- It may not be a profitable enterprise because AI-generated work is not currently copyrightable, so the producer wouldn’t really “own” anything over which they would have exclusive control (NOTE: This fact can be changed by revisions to the copyright law or bad judicial opinions, and there is already a blurry line between AI-generated work and work created by a human using AI as a tool);
- There is potential legal liability for producing AI-generated work, since it is based on material used without permission and the resulting play may end up being substantially similar to one of the ingested works, and so a producer would be unable to determine whether they were producing a work that was infringing someone’s copyright. Also, even where no substantial similarity exists, works generated “in the style of” another writer may violate trademark laws instead (NOTE: Unless the law is clarified to eliminate this potential liability or they can obtain insurance to cover it.;
- Unlike film and TV, or trade paperbacks, or news reports, or visual arts, theater is not a “mass medium” that people consume in bulk, including online. It is a niche artform, consumed at a relatively high price for educated and discerning audiences who are attending in person. The technology isn’t yet capable of producing commercially viable dramatic work that can create anything that would interest such an audience in spending what it currently costs to attend theater. (NOTE: No one ever went broke underestimating the tastes of the American public, so this, too, may not be an impediment over time if audiences are willing to buy tickets to it anyway); and
- “AI” is not creative; it’s predictive. It is based on large language models (LLMs) and generates content in response to specific queries based on its analysis of what it has already ingested and what is likely to be the next “most likely” response in a sequence. So, it will still require creative people to invent the queries, select among various responses, and rewrite and edit the resulting output. These people are called “dramatists”, so producers will need to hire writers anyway (NOTE: Until they don’t).
3. What is stopping publishing and licensing houses from using AI to create short plays for schools? In my experience in licensing, I have found that when looking for one-act plays, schools often pay more attention to things like cast-size, length, themes, etc., and focus less on who wrote a particular play. Knowing this, will publishers use AI to create these types of plays?
Ralph Sevush: Quite possibly. As stated previously, a less discerning audience can be exploited in this way. However, there are still the other legal and cultural obstacles described above. There is also legislation being proposed that would require AI-generated work to be labeled as such, and governmental institutions (like schools) may be legally prohibited from licensing such labeled work. Even if not prohibited, such work could be stigmatized by such labeling, limiting its value.
4. Dramatists aren’t the only creatives whose livelihood is potentially at risk. How will theatres and producers use AI to save money on design work? From sets to lighting to costumes, what is stopping a producer from using AI to create designs?
Ralph Sevush: United Scenic Artists is a labor union representing designers and so they can protect themselves by making it the subject of collective bargaining, like the WGA and SAG/AFTRA did. Dramatists have no such protections.
5. What is the ethical obligation for dramatists who want to use AI? Should writers have to disclose when their work was aided by AI?
Ralph Sevush: If AI is used by writers as a tool, then it raises no more ethical concerns than does using a word processor instead of a typewriter (or a quill pen and parchment). But, as stated above, the use of AI is not binary; it exists on a continuum. Using a program to help you brainstorm ideas for possible dramatization is one thing; leaving it to a computer program to write dialogue, delineate characters, and make the host of creative choices that the writer must make in crafting a play undermines and devalues the craft of writing and its practitioners, who’ve devoted years in harnessing their talents and honing their skills.
If you’re putting in some queries and then generating numerous works for submissions to theaters, festivals, contests, agents and publishers, who are then overwhelmed with AI-generated scripts for consideration, then you should most certainly be labeling your work as such, so such groups can filter you out of their selection process.
Okay, I lied again. I’m sorry. But there are many more questions we all need to be asking about the impact of AI on theatre. One more question I would like to propose: How do we, as a theatrical community, begin to install guardrails to prevent AI from limiting our earnings, and degrading our art? From publisher contracts to production agreements, we all need to be thinking about language that we can incorporate to protect the livelihood of all theatrical workers. And if any of this keeps you up at night, blame the robots, not me.